M6-013Medium模拟与校验 · Simulation & validation
如何用模拟向面试官解释“为什么 p 值不是 H0 为真的概率”?
Use a simulation to explain why a p-value is not the probability that the null is true.
💡 考场高分一句话答案:
模拟:在 H0 为真时重复实验,统计 p<0.05 的比例 ≈ 5%;再展示 p 值的分布是均匀的,与 H0 真假无关。
Simulate under a true null, repeat ten thousand times, and show that the p-value distribution is uniform and that about 5% fall below 0.05.
📌 核心考察要点 (Key Points)
- 模拟设计:H0 为真(两组同分布),重复 10000 次,画 p 值直方图
- 结论:p 值在 H0 下均匀分布,因此 p=0.03 不代表“H0 有 3% 概率为真”
- 对比:用贝叶斯后验才能给 $P(H_0\mid D)$
- 价值:这个模拟能在 30 秒内说服非技术听众
- Design: both arms drawn from the same distribution, so the null is true by construction
- Result: the p-value histogram is uniform, so p=0.03 does not mean a 3% chance the null holds
- Contrast: only a Bayesian posterior gives $P(H_0\mid D)$
- Value: this simulation persuades a non-technical audience in thirty seconds
🔍 大厂面试官连环追问 (Follow-ups)
- 如果模拟出的假阳性率是 8%,说明什么?
- 如何模拟“功效不足”的情形?
嫌字多看太累?进入苏格拉底式对话推演实验室
零打字负担,以大厂考官真实困境切入,通过点选直觉选项暴露思维盲区,手把手推导 Why。
⚡
Test muscle memory on this topic via 3-Layer Flashcards
Active Recall: Mentally formulate Prompt ➔ Math Derivation ➔ Benchmark Implementation in 3 minutes.